Hey readers!
I'm Romir - in short, I'm a undergrad at Carnegie Mellon University, studying Artificial Intelligence within the School of Computer Science. I am also currently working full-time as a Member of Technical Staff at Bespoke Labs!
I've also done research coast-to-coast in the U.S., working on many cool things: from PINNs to Diffusion-text models and beyond!
Age 0-13
I grew up in the Bay Area, close to San Jose in California. For my entire life there, I attended a private school, where I studied many things you would expect in an elementary and middle-school education. At this point, I wanted to go into medicine quite strongly - in fact, when I was 7, I developed a passionate hate for programming - most likely due to spending many hours of each day during summer in a Python course.
I had actually "technically" started coding since first grade - which was in a language called Terrapin Logo. The purpose of the language was simple: draw cool stuff. As much as I wish I could say "I built a website in first grade", the language just couldn't do anything nearly as complex. I did program a drawing of a sunflower though!
I started really picking up programming when I entered in eight grade(wild to think I was 12 then!). We finally progressed for Terrapin Logo to Python - an infinitely more useful language. And I actually started to enjoy programming around this time.
Around the end of eight grade - I moved to Arizona - a totally new state with (very!) different weather.
Age 13-17
When I entered high-school, I'm not sure as to why, but I knew I just wanted to do more. Now what that more was - I didn't know. After all - how is someone this young supposed to know?
But, I did know that the general direction I wanted to go in was "to help people". Very original - I know. So - to that end I created a club - "Helping Hands". We were somewhat similar to VolunteerMatch(a platform connecting facilities to students), but we actually helped with the onboarding process - and sometimes even helped to establish volunteering programs from scratch.
I did this exclusively for a year and a few extra months - through which I raised grants of $500,000, worked with hundreds others, and had a blast overall.
But then I realized that I was still interesting in programming(I had taken Python as the only freshman in the course, and Intro to CS in my first year). So - I wanted to work on something in the real-world(note: this was significantly before Claude Code/Codex went mainstream). I sent hundreds of emails to professors at Arizona State University - yielding...2 responses. One professor who went on to ghost me - and another who allowed me to work with him(Professor Tejaswi Gowda).
I worked on this towards the end of my sophomore year, and throughout my junior year as well. We were building a motion-capture suit fully from scratch. Not just the hardware - but the software behind it as well. We also ended up publishing a paper on this(arXiv), and achieving a level of accuracy nearly matching systems significantly more expensive.
Concurrently - I "shot my shot" at landing research at even more universities. My target was really aimed at one: MIT. After (way more) emails than before, I ended up getting two maybes: one from a Ph.D. student in the Raines group, and one from a Research Scientist II at the Braatz Lab.
I was able to start working with the Ph.D. student fairly quickly: I was working on modeling simulations related to anti-CRISPR proteins. However, before being able to work with the Research Scientist, there were many hurdles.
Throughout a few weeks - I had to read papers(significantly) above my proficiency. Think partial-differential equations, Bayes' Theorem, etc... and recreate the results of certain papers. While this was first when I started staying up extraordinarily late, it was also incredibly fun.
I did end up being able to work within the Braatz group - leading to my time at MIT over Summer 2025. I was there from June until July - where I worked on PINNs with PCE with MPC. Over the 6-7 weeks I was there, I pulled my first all nighter(x6-8 times). Traditionally, I stayed at the lab from 10am until 10pm(of course, unless I didn't go back for the night). I will admit: Domino's delivery was a life-saver during this time.
While I was in-person at MIT, I wanted to be more involved with AI. I had been programming in Python for a while unto this point, and Machine Learning certainly didn't seem to be fleeting. So, after a few emails, I was connected to a lab at MIT's CSAIL, along with another lab at the MIT McGovern Institute. While both of the projects were around post-training, they were in significantly different areas: one for formal theorem proving, and the other for predicting psychosocial conditions. This also resulted in an acceptance to NeurIPS Math-AI, and a conditional acceptance to Nature.
The day I flew back, as I was walking into the Cheesecake Factory, I was finishing up a call with a Ph.D. student at Caltech in Professor Anima Anandkumar's lab. While this was an incredibly surprising call to have been on, most working on research across 3 MIT departments and going through the most critical year of high school would have said no. But something I've learned is that I really, REALLY hate giving up opportunities. So: I said yes.
There, I worked on finetuning Diffusion-Text LLMs for proving theorems to answer the question: can diffusion-text LLMs match(or exceed) the performance of traditional LLMs, while beating them on speed?
Then, throughout my senior year, I was INCREDIBLY swamped with writing essays for the many universities I ended up applying to. This "busy season" lasted until about the end of December/start of January.
However, also at the end of December, I ended up signing to be an intern at AgentHub(now Panoptive), a startup backed by Y Combinator in the Summer 2025 batch. I ended up being their first ever hire - working on building their synthetic data generation pipeline end-to-end. This was my first taste of industry: and it was GOOD.
Towards the end of my internship, I also got many acceptances, waitlists, and rejections from the many universities I had applied to. After receiving my final decision at the start of April(which in retrospect wouldn't have swayed my decision at all, so there was no point in waiting), I ended up choosing Carnegie Mellon's School of Computer Science.
My internship then came to an end in March. The internship was part-time, at 20 hours a week, so it was something easily manageable.
On the last week of my internship, I got a LinkedIn message from a recruiter: there's a startup that could be interesting for you: would you be interested to hop on a call? Of course, I said "Why not?" and did so. I ended up shortly thereafter being on an introduction call with Mahesh Sathiamoorthy, the CEO of Bespoke Labs. The first question he asked was what I wanted: to be an intern? full-time? dropout?
The safe option then of course was to be an intern. And that's what I chose. After finishing the rest of the interview process at Bespoke, I ended up getting the offer to join: which I promptly took on.
From March/April until July, I was an intern at Bespoke, where I was working on building RL environments for frontier labs, neolabs, and enterprises. Yet, there was always a sort of tugging towards post-training: what I had spent years building towards.
In July - I made one of the most bold decisions I have (probably ever) made: I wrapped up at Bespoke as an intern. I joined them full-time, as their (youngest) Member of Technical Staff, as a Research Engineer.
Now: the obvious question that I got many, many texts about that day is: did you drop out? The answer is no. Remember how I said I really couldn't let go of opportunities I really enjoyed? To say Bespoke is one of them would be a pretty big understatement. For the past few weeks at Bespoke, I've been focusing on post-training, and it is genuinely the definition of "when you do what you love, you will never work a day in your life".
Now, I'm juggling just three things: full-time undergrad @ CMU in AI, in their School of CS + being a full-time MTS at Bespoke Labs + doing some (light) research at MIT. Not too much...right?
I also plan to start working on research under one of the professors at the CMU SCS somewhat soon. Though, this is very up in the air.
I also plan to stay at Bespoke for the foreseeable future. And, I hope to continue working on even more post-training work and research!
So: that's a summary of the past 17 years of my life.